Clémentine Grethen

Clémentine Grethen

I'm a PhD student at IRIT in Toulouse, France, where I work on computer vision, with a focus on 3D reconstruction and image processing. More specifically, my research studies multi-view reconstruction and rendering techniques, motivated by challenges in autonomous navigation. I'm funded by an ESA (European Space Agency) OSIP grant and supervised by Géraldine Morin and Simone Gasparini.

Email / Scholar / Github

Clémentine Grethen

Latest News

09 2026 Paper accepted Capturing Appearance and Material Optical Properties for Vision-Based Navigation has been accepted for oral presentation at ESA GNC 2026.
08 2026 Paper accepted Distilling Geometric Foundation Models for Stereo 3D Reconstruction: A Cross-Domain Study, an extension of our ECCV 2026 work, has been accepted for publication in the Springer ACVPR book.
06 2026 Paper accepted Geometric Foundation Model Distillation for Efficient Lunar 3D Reconstruction has been accepted at ECCV 2026.
05 2026 Speaker at the European Lunar Surface Workshop, presenting the talk "Robust Methods to Simulate Photorealistic Views of the Moon".
03 2026 Paper accepted Lunar-G2R has been accepted for publication at ICPR 2026.
02 2026 I am serving as a Reviewer for ICIP 2026.
12 2025 Presented my work at GTMG (Geometric Modeling Working Group) in Paris.
02 2026 Our paper MoonAnything has been accepted at ACM MMSys 2026!.
01 2026 New preprint on ArXiv: Lunar-G2R, focusing on BRDF estimation [ArXiv].
12 2025 Presented my work at the CNRS 3D Vision Day in Paris.
07 2025 Paper on LunarStereo accepted to ICCV 3D-VAST 2025, Honolulu.
06 2025 Presented our paper on moon landing at Orasis 2025.
10 2024 Started my PhD at IRIT Toulouse.

Research

2026
	Geometric Foundation Model Distillation for Efficient Lunar 3D Reconstructio Clémentine Grethen, Florient Chouteau, Géraldine Morin, Simone Gasparini ECCV 2026 [Paper] / [Project Page] / We distill a large MASt3R lunar stereo model into up to 7× smaller students that keep most reconstruction accuracy, and show that encoder strength and feature-level distillation are key for efficient 3D reconstruction under tight hardware constraints.
	MoonAnything: A Vision Benchmark with Large-Scale Lunar Supervised Data Clémentine Grethen, Yuang Shi, Simone Gasparini, Géraldine Morin ACM MMSys 2026 [Paper] / [Dataset] In this paper, we introduce MoonAnything, a unified benchmark built on real lunar topography with physically-based rendering, providing the first comprehensive geometric and photometric supervision under diverse illumination with large scale.
	Lunar-G2R: Geometry-to-Reflectance Learning for High-Fidelity Lunar BRDF Estimation Clémentine Grethen, Nicolas Menga, Roland Brochard, Simone Gasparini, Géraldine Morin, Jérémy Lebreton, Manuel Sanchez-Gestido ICPR 2026 (oral presentation) [Paper] / [Project Page] / [Dataset] We propose Lunar-G2R, a geometry-to-reflectance learning framework for high-fidelity BRDF estimation of the lunar surface, leveraging large-scale supervised data under diverse illumination and viewing conditions.
2025
	Adapting Stereo Vision From Objects To 3D Lunar Surface Reconstruction with the StereoLunar Dataset Clémentine Grethen, Simone Gasparini, Géraldine Morin, Jérémy Lebreton, Lucas Marti, Manuel Sanchez-Gestido ICCV workshop 2025 [Paper] / [Project Page] / [StereoLunar Dataset] We introduce LunarStereo, a photorealistic stereo image dataset of the Moon, and show that fine-tuning the MASt3R model enables accurate 3D reconstruction and pose estimation under lunar conditions, significantly outperforming baselines.
	Mise en place d'une méthode de reconstruction 3D du sol lunaire à partir de plusieurs images dans un contexte d'atterrissage. Clémentine Grethen, Simone Gasparini, Géraldine Morin Orasis, 2025 [HAL Paper] The paper proposes a method for generating realistic images of the lunar surface and a method for 3D reconstruction of the lunar surface based on traditional techniques.

This website is inspired from Jon Barron's. Last updated July 2025.